It's the effect of previous result[S] that arrived which had actual short processing times projection, possibly HCMD2 tasks, and then receiving one that had a short estimate, say 6 minutes, but took 6 hours (360 minutes). The client then assumes all other tasks of WCG also must be running 60 times longer ( 360 / 6). If you then had a half day work buffer, the client assumes that work will take 60 time longer too, so often clients start running in High Priority.

The unfortunate situation is, that it's impossible for the servers to give accurate estimates on a per-task basis, so they use the current actual average of recent day results and give that estimate to any new tasks going out.

The situation will balance itself over time. BOINC will correct the forward estimates from newly completed tasks when they run shorter, but that downward adjustment is slow [by design]. This prevents a client from fetching more work in case more tasks should arrive that get a wrong time estimate assignment of 6 minutes, then actually taking 6 hours.

Just let it run and in 20-30 more results, the estimates should be a lot better.

Question is: Did you have any tasks other than HCMD2 that ran very long? A stuck work unit could have caused this too, taking e.g. 24 hours wallclock time, but really only using 1 hour CPU time.

--//--

P.S. High buffer settings makes the situation worse as when you have e.g. 2 days cached and such a 60 times longer running result arrives, you suddenly have a buffer estimate of 120 days. All tasks will start processing in High Priority until the situation has normalized.

edit: And, welcome to the forums... 6 years after joining WCG. The majority of members never post a query... running the agent in set and forget mostly and let BOINC do it's thing :D